Real-time deep learning of robotic manipulator inverse dynamics

A.S. Polydoros; L. Nalpantidis; V. Kruger

doi:10.1109/IROS.2015.7353857

Real-time deep learning of robotic manipulator inverse dynamics

A.S. Polydoros, L. Nalpantidis, V. Kruger

Aalborg University

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

Abstract

In certain cases analytical derivation of physics-based models of robots is difficult or even impossible. A potential workaround is the approximation of robot models from sensor data-streams employing machine learning approaches. In this paper, the inverse dynamics models are learned by employing a novel real-time deep learning algorithm. The algorithm exploits the methods of self-organized learning, reservoir computing and Bayesian inference. It is evaluated and compared to other state of the art algorithms in terms of generalization ability, convergence and adaptability using five datasets gathered from four robots. Results show that the proposed algorithm can adapt to real-time changes of the inverse dynamics model significantly better than the other state of the art algorithms.

Original language	English
Title of host publication	2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS),
Place of Publication	United States
Publisher	IEEE - Institute of Electrical and Electronics Engineers Inc.
Pages	3442-3448
Number of pages	7
ISBN (Electronic)	978-1-4799-9994-1
DOIs	https://doi.org/10.1109/IROS.2015.7353857
Publication status	Published - 2015 Sept 28
Externally published	Yes
Event	IEEE/RSJ International Conference on Intelligent Robots and Systems , 2015 - Hamburg, Hamburg, Germany Duration: 2015 Sept 28 → 2015 Oct 2

Conference

Conference	IEEE/RSJ International Conference on Intelligent Robots and Systems , 2015
Abbreviated title	IROS2015
Country/Territory	Germany
City	Hamburg
Period	2015/09/28 → 2015/10/02

Free keywords

approximation theory
belief networks
inference mechanisms
learning (artificial intelligence)
real-time systems
Bayesian inference
machine learning
physics-based models
real-time deep learning
reservoir computing
robotic manipulator
inverse dynamics
self-organized learning
sensor data-streams
Adaptation models
computational modeling
Heuristic algorithms
Manipulator dynamics
Robot sensing systems

Access to Document

10.1109/IROS.2015.7353857

Cite this

@inproceedings{2dcf04e9f81242d8b14c7a804a0dd187,

title = "Real-time deep learning of robotic manipulator inverse dynamics",

abstract = "In certain cases analytical derivation of physics-based models of robots is difficult or even impossible. A potential workaround is the approximation of robot models from sensor data-streams employing machine learning approaches. In this paper, the inverse dynamics models are learned by employing a novel real-time deep learning algorithm. The algorithm exploits the methods of self-organized learning, reservoir computing and Bayesian inference. It is evaluated and compared to other state of the art algorithms in terms of generalization ability, convergence and adaptability using five datasets gathered from four robots. Results show that the proposed algorithm can adapt to real-time changes of the inverse dynamics model significantly better than the other state of the art algorithms.",

keywords = "approximation theory, belief networks, inference mechanisms, learning (artificial intelligence), real-time systems, Bayesian inference, machine learning, physics-based models, real-time deep learning, reservoir computing, robotic manipulator, inverse dynamics, self-organized learning, sensor data-streams, Adaptation models, computational modeling, Heuristic algorithms, Manipulator dynamics, Robot sensing systems",

author = "A.S. Polydoros and L. Nalpantidis and V. Kruger",

year = "2015",

month = sep,

day = "28",

doi = "10.1109/IROS.2015.7353857",

language = "English",

pages = "3442--3448",

booktitle = "2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS),",

publisher = "IEEE - Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "IEEE/RSJ International Conference on Intelligent Robots and Systems , 2015, IROS2015 ; Conference date: 28-09-2015 Through 02-10-2015",

}

Polydoros, AS, Nalpantidis, L & Kruger, V 2015, Real-time deep learning of robotic manipulator inverse dynamics. in 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), . IEEE - Institute of Electrical and Electronics Engineers Inc., United States, pp. 3442-3448, IEEE/RSJ International Conference on Intelligent Robots and Systems , 2015, Hamburg, Germany, 2015/09/28. https://doi.org/10.1109/IROS.2015.7353857

Real-time deep learning of robotic manipulator inverse dynamics. / Polydoros, A.S.; Nalpantidis, L.; Kruger, V.
2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), . United States: IEEE - Institute of Electrical and Electronics Engineers Inc., 2015. p. 3442-3448.

Research output: Chapter in Book/Report/Conference proceeding › Paper in conference proceeding › peer-review

TY - GEN

T1 - Real-time deep learning of robotic manipulator inverse dynamics

AU - Polydoros, A.S.

AU - Nalpantidis, L.

AU - Kruger, V.

PY - 2015/9/28

Y1 - 2015/9/28

N2 - In certain cases analytical derivation of physics-based models of robots is difficult or even impossible. A potential workaround is the approximation of robot models from sensor data-streams employing machine learning approaches. In this paper, the inverse dynamics models are learned by employing a novel real-time deep learning algorithm. The algorithm exploits the methods of self-organized learning, reservoir computing and Bayesian inference. It is evaluated and compared to other state of the art algorithms in terms of generalization ability, convergence and adaptability using five datasets gathered from four robots. Results show that the proposed algorithm can adapt to real-time changes of the inverse dynamics model significantly better than the other state of the art algorithms.

AB - In certain cases analytical derivation of physics-based models of robots is difficult or even impossible. A potential workaround is the approximation of robot models from sensor data-streams employing machine learning approaches. In this paper, the inverse dynamics models are learned by employing a novel real-time deep learning algorithm. The algorithm exploits the methods of self-organized learning, reservoir computing and Bayesian inference. It is evaluated and compared to other state of the art algorithms in terms of generalization ability, convergence and adaptability using five datasets gathered from four robots. Results show that the proposed algorithm can adapt to real-time changes of the inverse dynamics model significantly better than the other state of the art algorithms.

KW - approximation theory

KW - belief networks

KW - inference mechanisms

KW - learning (artificial intelligence)

KW - real-time systems

KW - Bayesian inference

KW - machine learning

KW - physics-based models

KW - real-time deep learning

KW - reservoir computing

KW - robotic manipulator

KW - inverse dynamics

KW - self-organized learning

KW - sensor data-streams

KW - Adaptation models

KW - computational modeling

KW - Heuristic algorithms

KW - Manipulator dynamics

KW - Robot sensing systems

U2 - 10.1109/IROS.2015.7353857

DO - 10.1109/IROS.2015.7353857

M3 - Paper in conference proceeding

SP - 3442

EP - 3448

BT - 2015 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS),

PB - IEEE - Institute of Electrical and Electronics Engineers Inc.

CY - United States

T2 - IEEE/RSJ International Conference on Intelligent Robots and Systems , 2015

Y2 - 28 September 2015 through 2 October 2015

ER -

Real-time deep learning of robotic manipulator inverse dynamics

Abstract

Conference

Free keywords

Access to Document

Fingerprint

Cite this